US8813052B2 - Cache metadata for implementing bounded transactional memory - Google Patents

Cache metadata for implementing bounded transactional memory Download PDF

Info

Publication number
US8813052B2
US8813052B2 US11/811,370 US81137007A US8813052B2 US 8813052 B2 US8813052 B2 US 8813052B2 US 81137007 A US81137007 A US 81137007A US 8813052 B2 US8813052 B2 US 8813052B2
Authority
US
United States
Prior art keywords
cache
metadata
clmd
transaction
cache line
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active, expires
Application number
US11/811,370
Other languages
English (en)
Other versions
US20070245099A1 (en
Inventor
Jan Gray
Timothy L. Harris
James Larus
Burton Smith
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Microsoft Technology Licensing LLC
Original Assignee
Microsoft Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
US case filed in Texas Eastern District Court litigation Critical https://portal.unifiedpatents.com/litigation/Texas%20Eastern%20District%20Court/case/4%3A14-cv-00827 Source: District Court Jurisdiction: Texas Eastern District Court "Unified Patents Litigation Data" by Unified Patents is licensed under a Creative Commons Attribution 4.0 International License.
Priority claimed from US11/389,451 external-priority patent/US8799882B2/en
Priority to US11/811,370 priority Critical patent/US8813052B2/en
Application filed by Microsoft Corp filed Critical Microsoft Corp
Publication of US20070245099A1 publication Critical patent/US20070245099A1/en
Assigned to MICROSOFT CORPORATION reassignment MICROSOFT CORPORATION ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: GRAY, JAN, HARRIS, TIMOTHY L., LARUS, JAMES, SMITH, BURTON
Priority to TW097117848A priority patent/TWI498733B/zh
Priority to PCT/US2008/065376 priority patent/WO2008154191A2/en
Priority to EP08769911.2A priority patent/EP2174223B1/de
Publication of US8813052B2 publication Critical patent/US8813052B2/en
Application granted granted Critical
Assigned to MICROSOFT TECHNOLOGY LICENSING, LLC reassignment MICROSOFT TECHNOLOGY LICENSING, LLC ASSIGNMENT OF ASSIGNOR'S INTEREST Assignors: MICROSOFT CORPORATION
Active legal-status Critical Current
Adjusted expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/466—Transaction processing
    • G06F9/467—Transactional memory
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F12/00—Accessing, addressing or allocating within memory systems or architectures
    • G06F12/02—Addressing or allocation; Relocation
    • G06F12/08—Addressing or allocation; Relocation in hierarchically structured memory systems, e.g. virtual memory systems
    • G06F12/0802—Addressing of a memory level in which the access to the desired data or data block requires associative addressing means, e.g. caches
    • G06F12/0806—Multiuser, multiprocessor or multiprocessing cache systems
    • G06F12/0808—Multiuser, multiprocessor or multiprocessing cache systems with cache invalidating means
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F12/00—Accessing, addressing or allocating within memory systems or architectures
    • G06F12/02—Addressing or allocation; Relocation
    • G06F12/08—Addressing or allocation; Relocation in hierarchically structured memory systems, e.g. virtual memory systems
    • G06F12/0802—Addressing of a memory level in which the access to the desired data or data block requires associative addressing means, e.g. caches
    • G06F12/0806—Multiuser, multiprocessor or multiprocessing cache systems
    • G06F12/0842—Multiuser, multiprocessor or multiprocessing cache systems for multiprocessing or multitasking
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
    • G06F9/30003—Arrangements for executing specific machine instructions
    • G06F9/3004—Arrangements for executing specific machine instructions to perform operations on memory
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
    • G06F9/30003—Arrangements for executing specific machine instructions
    • G06F9/3004—Arrangements for executing specific machine instructions to perform operations on memory
    • G06F9/30047—Prefetch instructions; cache control instructions
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
    • G06F9/30003—Arrangements for executing specific machine instructions
    • G06F9/30076—Arrangements for executing specific machine instructions to perform miscellaneous control operations, e.g. NOP
    • G06F9/30087—Synchronisation or serialisation instructions
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
    • G06F9/38—Concurrent instruction execution, e.g. pipeline or look ahead
    • G06F9/3824—Operand accessing
    • G06F9/3834—Maintaining memory consistency
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
    • G06F9/38—Concurrent instruction execution, e.g. pipeline or look ahead
    • G06F9/3854—Instruction completion, e.g. retiring, committing or graduating
    • G06F9/3858—Result writeback, i.e. updating the architectural state or memory
    • G06F9/38585—Result writeback, i.e. updating the architectural state or memory with result invalidation, e.g. nullification
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2212/00—Indexing scheme relating to accessing, addressing or allocation within memory systems or architectures
    • G06F2212/10—Providing a specific technical effect
    • G06F2212/1016—Performance improvement
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2212/00—Indexing scheme relating to accessing, addressing or allocation within memory systems or architectures
    • G06F2212/46—Caching storage objects of specific type in disk cache
    • G06F2212/466—Metadata, control data
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2212/00—Indexing scheme relating to accessing, addressing or allocation within memory systems or architectures
    • G06F2212/50—Control mechanisms for virtual memory, cache or TLB
    • G06F2212/507—Control mechanisms for virtual memory, cache or TLB using speculative control

Definitions

  • a CPU cache is a computer hardware mechanism used by the central processing unit of a computer to reduce the average time to access memory.
  • a cache is a small, fast memory that retains copies of the data from recently used main memory locations. If a subsequent memory access is to a memory address that has been retained in the cache, the memory access is satisfied using the cache memory. Thus, the more accesses that are performed from cached memory locations, the lower the average memory access time, and the faster the application program runs.
  • Cache memory is subdivided into cache lines. Each cache line has a copy of some fixed-size, contiguous range of bytes of main memory. Each cache line also has an address tag and other state that identifies whether the cache line is presently valid and if so what addressed range of data are retained in the cache line. Cache lines are of a fixed size, typically 32 to 256 bytes, that depends upon the hardware design.
  • a CPU When a CPU performs a read or write memory access to a data at a given address in main memory, it also checks whether that address is contained in its cache, in particular, if the cache contains a cache line which is valid and whose address tag matches the memory address of the access. If so, then a cache hit occurs, and the CPU accesses the data in the cache line.
  • a cache miss occurs and the CPU proceeds down the slower path of accessing the data elsewhere, and recording a copy of the data in a cache line in the CPU cache. Since a cache is of fixed size, to retain new data in a cache line, it may be necessary to evict (invalidate) data that may have been previously held in that cache line.
  • Software application programs execute a sequence of hardware instructions to effect a computation. Such instructions may perform arithmetic operations, may alter the program control flow sequence of subsequent instructions, may read or write (collectively, access) data at specific memory addresses, or perform other operations.
  • a CPU cache is used with a CPU, its presence, and any information about which addresses are currently cached, and any hardware processes of checking, validating, and invalidating cache lines, is typically invisible and inaccessible to software programs, except that programs usually execute faster.
  • Modern computers may have a plurality of layers of caches. For example, a small, fast level one cache (L1$) may quickly service most memory accesses; but on an L1$ miss, a larger, slower level two cache (L2$) may be accessed. Only upon a memory access that misses both the L1$ and L2$ would the access be performed to the relatively very slow main memory.
  • L1$ fast level one cache
  • L2$ slower level two cache
  • Modern computers may also be multiprocessors, which have a plurality of CPUs.
  • each CPU may access the same shared memory, so one CPU may write to shared memory and later another CPU may read the data written by the first.
  • Each CPU may have one or more layers of cache for its exclusive use (private cache) as well as one or more layers of caches shared with other CPUs (shared cache).
  • cache coherence protocols implement cache coherence to transparently provide the multiple threads of execution in the software program with the illusion that all memory accesses are to a single common shared main memory.
  • cache coherence protocols Sometimes an access in one CPU (such as a write) must invalidate a cache line in other CPUs.
  • a logical plurality of CPUs can be implemented in hardware in a time multiplexed fashion upon a single CPU core, by providing a plurality of copies of all the processor state and registers, called hardware thread contexts, in a single CPU.
  • This is known as a multithreaded CPU core.
  • a single CPU core with four distinct thread contexts e.g.
  • the metadata state can include at least some bits of state for each quantum of addressed data, at least some state for each cache line, and at least some state for the cache overall. Additional instructions in the central processing unit are provided for interacting with this metadata. New side effects are introduced into operations of the central processing unit and cache by presence of the metadata and the additional instructions.
  • the metadata can be accessed by at least one software program to facilitate an operation of the software program.
  • a bounded transactional memory application accesses cache metadata in a cache of a central processing unit.
  • a cache line metadata transaction-read bit is set.
  • a cache line metadata transaction-write bit is set and a conditional store is performed.
  • the application can also interrogate a cache line metadata eviction summary to determine whether a transaction is doomed and then take an appropriate action.
  • a hardware accelerated software transactional memory (HASTM) application has access to metadata in a cache of a central processing unit that can be used to improve the operation of the STM system, in particular to accelerate some of the most time consuming operations in the software transactional memory operation.
  • HASTM hardware accelerated software transactional memory
  • the software transactional memory application has access to metadata in a cache of a central processing unit that can be used to improve the operation of the STM system, in particular to accelerate some of the most time consuming operations in the software transactional memory operation.
  • open read barrier filtering is provided that uses an opened-for-read bit that is contained in the cache metadata to quickly filter (test and set) whether a given transactional memory datum has already received the expensive software bookkeeping it requires. If so, the redundant ‘open for read’ software bookkeeping is safely skipped.
  • Read log validation is also accelerated using metadata.
  • the HASTM software sets a read-set bit in the cache line metadata for each accessed datum, accumulating in the cache a ‘read set’ of metadata state representing the set of data the transaction read.
  • This cache metadata is retained alongside its cached data unless evicted or unless invalidated when some other CPU writes to data in this CPU's read set.
  • the application may interrogate the cache line metadata eviction summary; if no lines were evicted, the CPU's read set is intact, so software read set validation is unnecessary and can be skipped.
  • This same read-set cache line metadata also accelerates the retry facility of a software transactional memory system.
  • the HASTM software need only establish a software handler vector to await invalidation of a read-set bit of cache line metadata. Then when another CPU writes some data that is in the application's read set, the corresponding cache line (and hence its read-set bit in its cache line metadata) will be invalidated, triggering a jump to a software handler that concludes the ‘retry’ action and resumes (restarts) the transaction.
  • FIG. 1 is a diagrammatic view of a computer system of one implementation.
  • FIG. 2 is a more detailed diagrammatic view of a central processing unit of one implementation operating on the computer system of FIG. 1 .
  • FIG. 3 is a diagram illustrating exemplary hardware structure that implements additional metadata per logical processor per cache line for the system of FIG. 1 .
  • FIG. 4 is a diagram illustrating exemplary hardware structure that implements additional metadata per logical processor per cache for the system of FIG. 1 .
  • FIG. 5 is diagrammatic view of a central processing unit of the system of FIG. 1 illustrating an exemplary instruction set architecture and its interaction with cache metadata.
  • FIG. 6 is a diagrammatic view of exemplary automatic cache and processor operation instructions for the central processing unit of FIG. 5 .
  • FIG. 7 is a process flow diagram for one implementation of the system of FIG. 1 that illustrates the stages involved in loading a cache line and initializing some cache metadata to default values.
  • FIG. 8 is a process flow diagram for one implementation of the system of FIG. 1 that illustrates the stages involved in evicting or invalidating a cache line.
  • FIG. 9 is a diagrammatic view of exemplary CMD instructions for the central processing unit of FIG. 5 .
  • FIG. 10 is a diagrammatic view of exemplary VAMD instructions for the central processing unit of FIG. 5 .
  • FIG. 11 is a diagrammatic view of exemplary CLMD instructions for the central processing unit of FIG. 5 .
  • FIG. 12 is a diagrammatic view of exemplary context switch save and restore extensions for the central processing unit of FIG. 5 .
  • FIG. 13 is a diagrammatic view of a bounded transactional memory application of one implementation operating on the computer system of FIG. 1 .
  • FIG. 14 is a process flow diagram for one implementation of the system of FIG. 1 that illustrates the stages involved in providing a bounded transactional memory application using cache metadata.
  • FIG. 15 is a process flow diagram for one implementation of the system of FIG. 1 that illustrates the stages involved in using the get CLMD evictions instruction to poll whether a transaction is doomed.
  • FIG. 16 is a process flow diagram for one implementation of the system of FIG. 1 that illustrates the stages involved in using an addition to the CMD structure to handle doomed transactions in hardware.
  • FIG. 17 is a diagrammatic view of a hardware accelerated software transactional memory application of one implementation operating on the computer system of FIG. 1 .
  • FIG. 18 is a process flow diagram for one implementation of the system of FIG. 1 that illustrates the stages involved in providing Open_Read barrier filtering that uses the opened-for-read bit on the VAMD of the CPU cache to avoid redundant filtering.
  • FIG. 19 is a process flow diagram for one implementation of the system of FIG. 1 that illustrates the stages involved in providing Open_Write barrier filtering that uses the TMW bit on the CLMD of the CPU cache.
  • FIG. 20 is a process flow diagram for one implementation of the system of FIG. 1 that illustrates the stages involved in providing Write_Undo barrier filtering that uses the logged-for-undo bit to efficiently filter out redundant undo logging.
  • FIG. 21 is a process flow diagram for one implementation of the system of FIG. 1 that illustrates the stages involved in providing read log validation that uses the get CLMD evictions instruction on the CPU to avoid unnecessary read log validations.
  • FIG. 22 is a process flow diagram for one implementation of the system of FIG. 1 that illustrates the stages involved in providing retry operations that mark CLMD lines.
  • the system may be described in the general context as a system that has a central processing unit that uses cache metadata on the CPU cache for improving the operation of one or more software programs.
  • an exemplary computer system to use for implementing one or more parts of the system includes a computing device, such as computing device 100 .
  • computing device 100 In its most basic configuration, computing device 100 typically includes at least one processing unit 102 and memory 104 .
  • central processing unit(s) 102 include a cache 103 with software accessible metadata 105 . These metadata are described in further detail in several other figures herein.
  • One or more hardware assisted software application(s) 150 can access the metadata 105 of the CPU cache 103 to facilitate an operation of the respective software application.
  • hardware assisted software applications can include, but are not limited to transactional memory systems, garbage collection systems, systems for analyzing the performance or run-time behavior of programs, systems for finding defects in programs, systems for enforcing security constraints on programs, and/or any other types of software applications that can be improved (either in performance or in some other way) by using the metadata 105 on the CPU cache 103 .
  • These software applications may directly read and write the metadata as appropriate for the given scenario.
  • the hardware itself may read and/or modify the metadata as appropriate.
  • memory 104 may be volatile (such as RAM), non-volatile (such as ROM, flash memory, etc.) or some combination of the two. This most basic configuration is illustrated in FIG. 1 by dashed line 106 .
  • device 100 may also have additional features/functionality.
  • device 100 may also include additional storage (removable and/or non-removable) including, but not limited to, magnetic or optical disks or tape.
  • additional storage is illustrated in FIG. 1 by removable storage 109 and non-removable storage 110 .
  • Computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data.
  • Memory 104 , removable storage 109 and non-removable storage 110 are all examples of computer storage media.
  • Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can be accessed by device 100 . Any such computer storage media may be part of device 100 .
  • Computing device 100 includes one or more communication connections 114 that allow computing device 100 to communicate with other computers/applications 115 .
  • Device 100 may also have input device(s) 112 such as keyboard, mouse, pen, voice input device, touch input device, etc.
  • Output device(s) 111 such as a display, speakers, printer, etc. may also be included. These devices are well known in the art and need not be discussed at length here.
  • FIG. 2 is a more detailed diagrammatic view of a central processing unit 102 of one implementation operating on the computer system of FIG. 1 .
  • Cache 103 of CPU 102 includes software accessible cache metadata 105 .
  • This metadata (for short) is additional, software accessible state associated with cached data.
  • the cache metadata state is retained only for the time interval that its corresponding data is cached.
  • the software accessible metadata 105 include per virtual address metadata (called VAMD herein) 106 , per cache line metadata (called CLMD herein), and/or per cache metadata control registers (called CMD herein) 108 .
  • the VAMD 106 , CLMD 107 , and/or CMD 108 metadata in the CPU cache 103 can be accessed by one or more software applications to improve an operation of the respective application.
  • some, additional, and/or other types of metadata could be used in the cache than those shown in FIG. 2 , or in a hardware location other than the cache.
  • the VAMD state to be described could reside in a separate cache-like structure disjoint from CPU cache(s)
  • the CLMD state could reside in a separate centralized or distributed structure that interacts with the cache coherence system but is nevertheless disjoint from CPU cache(s).
  • these metadata 105 will be discussed in further detail in various figures herein to describe some techniques for improving software application operation.
  • FIG. 3 is a diagram illustrating exemplary hardware instructions 170 that implement additional metadata per logical processor per cache line for the system of FIG. 1 .
  • the term logical processor as used herein is meant to include each of the one or more actual CPU cores, and/or multithreaded CPU cores' hardware thread contexts ([007]) that share a single cache.
  • each virtual address metadata (VAMD) of each quadword of the cache line of each logical processor is allocated four bits of metadata
  • each cache line metadata (CLMD) of each cache line of each logical processor is allocated eight bits of metadata.
  • VAMD virtual address metadata
  • CLMD cache line metadata
  • a VAMD is assigned to each 64-bit quadword of data in the cache line.
  • a quadword is because it is currently the smallest power-of-two block size for which there will never be two separate transacted objects in the same block.
  • something smaller or larger than a quadword could be used for each VAMD and still take advantage of some or all of the various techniques discussed herein. It is important to note there is but one CLMD per cache line per logical processor that shares the cache; but there are potentially many VAMDs (one quadword) per cache line per logical processor.
  • FIG. 4 a diagram is shown that illustrates exemplary hardware state 190 that implement cache metadata control registers per logical processor per cache.
  • these registers 190 control and track various cache metadata behaviors, including CLMD evictions summaries, CLMD speculative writes, a CLMD default value, a VAMD default value, a transaction handler vector address and an CLMD eviction mask that triggers a transaction handler invocation.
  • CLMD evictions summaries including CLMD evictions summaries, CLMD speculative writes, a CLMD default value, a VAMD default value, a transaction handler vector address and an CLMD eviction mask that triggers a transaction handler invocation.
  • CLMD evictions summaries including CLMD evictions summaries, CLMD speculative writes, a CLMD default value, a VAMD default value, a transaction handler vector address and an CLMD eviction mask that triggers a transaction handler invocation.
  • numerous other allocations and/or arrangements of CPU cache metadata could be used than shown in this hypothetical example.
  • FIG. 5 a diagrammatic view of a central processing unit 200 is shown that illustrates an exemplary ‘cache metadata enhanced hardware instruction set architecture 202 provided by central processing unit 200 , and its interaction with cache metadata 224 of the CPU cache 222 .
  • instruction set architecture 202 includes automatic cache and processor operation behaviors 204 , CMD instructions 206 , VAMD instructions 208 , CLMD instructions 214 , and thread context switch save/restore extensions 220 .
  • the automatic cache and processor operation behaviors 204 can interact with various metadata 224 , as described in further detail in FIGS. 6-8 .
  • the CMD instructions 206 can interact with various metadata 226 , as described in further detail in FIG. 9 .
  • VAMD instructions 208 include both individual instructions 210 and flash instructions 212 .
  • the VAMD instructions 208 can interact with the VAMD metadata 228 and/or other metadata as described in further detail in FIG. 10 .
  • CLMD instructions 214 can interact with the CLMD metadata 230 and/or other metadata as described in further detail in FIG. 11 .
  • CLMD instructions 214 include both individual instructions 216 and flash instructions 218 .
  • the thread context switch save/restore extensions 220 can interact with various metadata 224 , as described in further detail in FIG. 12 .
  • FIG. 6 is a diagrammatic view of exemplary automatic cache and processor operation behaviors 250 for the central processing unit 200 of FIG. 5 .
  • These behaviors extend prior art CPU+cache operations so as to account for the new cache metadata state. This impacts initialization (validation) 252 , eviction/invalidation 258 , and core reset operations 270 .
  • Cache line initialization occurs implicitly upon a cache line 252 during the hardware processing of a cache miss.
  • Cache line initialization sets the line's CLMDs to the default value 254 , and its VAMDs to the default value 256 .
  • Below is an example hardware definition pseudocode that illustrates additional initialization behaviors 252 for setting the default CLMDs 254 and the default VAMDs 256 for each logical processor. These stages are described in more detail in FIG. 7 .
  • the eviction/invalidation operations 258 are designed to run when evicting or invalidating the cache line. Eviction occurs when a memory access from some logical processor on this cache forces some valid cache line to be repurposed to retain the newly accessed data. In that case, the data contents of the cache line are discarded or written back to memory, and the metadata contents are completely discarded. Invalidation occurs when a coherent memory access from another core forces a line to transition from a valid state to invalid in this particular level of the cache in this core.
  • An example hardware definition pseudocode is shown below for implementing these eviction/invalidation behaviors 258 . These are described in further detail in FIG. 8 .
  • Core reset instructions 270 can also be included to zero out all of the metadata 272 .
  • An example hardware instruction that zeros out all of the metadata is shown below. In the example shown, all cache metadata mode state bits are zeroed and all metadata is zeroed. For example, the CLMD evictions, CLMD speculative writes, the default CLMD value, and the default VAMD value are all set to zero.
  • FIG. 7 one implementation of the stages involved in loading a cache line and initializing some cache metadata to default values is illustrated in further detail (than as initially described as part of the initialization instructions 252 of FIG. 6 ).
  • the process of FIG. 7 is implemented in the hardware of computing device 100 .
  • the process begins at start point 290 with the CPU loading a cache line (stage 292 ).
  • the CLMD metadata is initialized to its logical-processor-specific default value (stage 294 ).
  • the VAMD metadata is similarly initialized to its LP-specific default value (stage 296 ).
  • the process ends at end point 300 .
  • FIG. 8 one implementation of the stages involved in evicting or invalidating a cache line is shown in further detail (than as initially described as part of the eviction/invalidation instructions 258 of FIG. 6 ).
  • the process of FIG. 8 is implemented in the hardware of computing device 100 .
  • the process begins at start point 310 with the CPU determining it is time to evict or invalidate a cache line (stage 312 ).
  • the CLMD's of the logical processor are propagated to the cache's eviction summary (stage 314 ).
  • the line's CLMD and VAMD bits are then discarded (stage 316 ). If the cache line was explicitly written speculatively, it is invalidated first (not written back to main memory) (stage 318 ).
  • the process ends at end point 320 .
  • FIG. 9 is a diagrammatic view of exemplary CMD instructions for the central processing unit 200 of FIG. 5 .
  • these instructions extend and complement a base set of hardware instructions provided by a conventional CPU. In other words, they extend the base instruction set architecture.
  • These hardware instructions are used by software to interact with various cache metadata state, and control cache metadata behaviors. For example, there are instructions for setting and getting the VAMD default value control register ( 332 and 334 ), and for setting and getting the CLMD default value control register ( 336 and 338 ).
  • An example of hardware instruction definition pseudocode that could be used to provide such functionality for setting and getting these default values is shown below.
  • the hardware definition pseudocode attribute “instruction” on a function designates the hardware operation is (or can be) made available as a novel CPU instruction for explicit use by software.
  • the function name suggests the CPU instruction name and its function arguments and return value suggest in- and out-parameters of the corresponding instruction, most commonly expressed as programmer specified registers or as implicit condition codes.
  • VAMD_DEFAULT // Set the current default VAMD control register for this logical processor.
  • instruction void cache_set_vamd_default(VAMD vamd) ⁇ dcache.cmds[lp].vamd_default vamd; ⁇ // GET_VAMD_DEFAULT: // Get the current default VAMD control register for this logical processor.
  • instruction void cache_set_clmd_default(CLMD clmd) ⁇ dcache.cmds[lp].clmd_default clmd; ⁇ // GET_CLMD_DEFAULT: // Set the current default CLMD control register for this logical processor.
  • CMD instructions 330 also include instructions for setting and getting the CLMD speculative writes control register ( 342 and 344 ), which sets and gets the control register that determines which CLMD bits that indicate the line has been speculatively written.
  • CLMD speculative writes control register 342 and 344
  • An example of hardware instruction definition pseudocode that can be used to set and get the speculative writes is shown below.
  • CMD instructions 330 includes instructions for setting and getting the CLMD evictions summary control register ( 344 and 346 ).
  • An example of hardware instruction definition pseudocode that can be used to set and get the CLMD evictions is shown below.
  • CMD instructions 330 includes instructions for setting and getting the CLMD eviction handler address and handler mask control registers ( 190 ).
  • An example of hardware instruction definition pseudocode that can be used to set and get the CLMD evictions is shown below.
  • instruction void cache_set_clmd_eviction_handler(CLMD clmd) ⁇ dcache.cmds[lp].clmd_eviction_handler_mask clmd; ⁇ // GET_CLMD_EVICTION_HANDLER_MASK: // Get the current CLMD evictions handler mask control register for this logical processor.
  • CMD instructions 330 can include hardware instructions for conditionally testing evictions with flash clearing and setting CLMD values on all cache lines 348 . This can be used in a bounded transactional memory system, or for other purposes, as described in further detail herein.
  • An example of hardware instruction definition pseudocode that can be used to conditionally test evictions with flash clear/set is shown below.
  • COND_TEST_EVICTIONS_AND_OR_ALL // Atomically test whether any specific CLMD bits' evictions or invalidations // have occurred; // and if not, flash clear (via AND) and flash set (via OR) specific CLMD bit positions.
  • CMD instructions 330 can include hardware instructions for conditionally discarding cache lines 349 based on the CLMD.
  • An example of hardware instruction definition pseudocode that can be used to conditionally discard cache lines is shown below.
  • CMD instructions can also include instructions for getting the implemented size of certain data, such as getting cache line size 350 , getting VAMD bits 352 , getting VAMD stride 354 , or getting CLMD size 356 .
  • An example hardware instruction definition pseudocode that can be used to get these basic metadata value sizes is shown below.
  • An alternative implementation might provide such implementation specific parameter data via an alternative mechanism such as a general-purpose CPUID instruction.
  • VAMD instructions 370 include individual instructions 372 that access specific VAMDs one at a time, and the flash instructions 388 that apply to all VAMDs in the cache.
  • the individual instructions 372 can include instructions for implementing a VAMD get 376 , a VAMD set 378 , a VAMD test and set 382 , a VAMD selective clear (and) 384 , and a VAMD selective set (OR) 386 .
  • VAMD helper pseudocode function that (like a read data instruction) takes an address ‘va’ and ensures its data is cached; then returns a reference (which in hardware, might be a control signal bit pattern) to the specific VAMD corresponding to the quadword of bytes at address va.
  • VAMD& vamd_va(VA va) validate_line(va); return dcache.lines[line_va(va)].vamds[offset_va(va)][lp]; ⁇
  • the VAMD get instruction 376 selects and returns the current value of the particular VAMD that is appropriate for the particular address.
  • the VAMD set instruction 378 stores a VAMD for the particular address.
  • Example hardware definition pseudocode instructions for the get and set instructions are shown below.
  • the VAMD test instruction 380 fetches the VAMD for the particular address, performs an AND operation with the VAMD and the mask, and compares the result. In most CPU's base instruction set architectures, such comparison results are typically written to condition code registers or to general purpose registers.
  • the VAMD test and set instruction 382 atomically tests and sets the address and then returns what was read before the set occurred. Example hardware instruction definition pseudocode for these two tests are shown below.
  • VAMD_TEST // Return true if all of the specified VAMD bits for the VAMD at ‘va’ are set.
  • instruction bool vamd_test_and_set(VAMD vamd, VA va) ⁇ atomic ⁇ bool ret vamd_test(vamd, va); vamd_or(vamd, va); return ret; ⁇ ⁇
  • the VAMD selective clear instruction 384 selectively clears the VAMD and the VAMD selective set instruction 386 selectively sets the VAMD, as further illustrated in the hardware instructions below.
  • VAMD_AND // Bitwise-AND the VAMD mask against the VAMD for the specified address ‘va’. // This may be used to clear certain VAMD bits.
  • VAMD vamd_and(VAMD vamd, VA va) ⁇ return vamd_va(va) & vamd; ⁇ // VAMD_OR // Bitwise-OR the VAMD mask against the VAMD for the specified address ‘va’.
  • vamd; ⁇
  • whole cache flash VAMD instructions 388 can be provided.
  • a flash clear (‘AND ALL’) instruction 390 can be provided, along with a flash set (‘OR_ALL’) instruction 392 .
  • the VAMD_AND_ALL instruction is designed to flash clear designated VAMD bits of all of the VAMDs for each cache line of this logical processor
  • the VAMD_OR_ALL instruction is designed to similarly flash set all of the VAMDs for each cache line of this logical processor.
  • VAMD_AND_ALL // Flash bitwise-AND the specified mask over all the current logical processor's // VAMDs.
  • vamd; ⁇
  • FIG. 11 a diagrammatic view of exemplary CLMD instructions 410 for the central processing unit 200 of FIG. 5 is shown.
  • the individual instructions can include instructions for CLMD get 416 , CLMD set 418 , CLMD test 420 , CLMD selective clear (AND) 422 , CLMD selective set (OR) 424 , and CLMD conditional store 426 .
  • These individual instructions operate similarly as just described with respect to the VAMD instructions, but are discussed herein briefly for the CLMD context. For example, the get and set instructions ( 416 and 418 ) get and set the value of the CLMD.
  • the test instruction fetches the CLMD for the particular address, performs an AND operation with the CLMD and the mask, and compares the result. The test returns true if the bits in the mask are all set.
  • the selective clear and selective set ( 422 and 424 ) performs selective clears or sets to the CLMD, respectively. Examples of hardware instruction definition pseudocode are shown below.
  • instruction void clmd_set(CLMD clmd, VA va) ⁇ clmd_va(va) clmd; ⁇ // CLMD_TEST // Return true if all of the specified CLMD bits for the CLMD for the specified address // ‘va’ are set.
  • the CLMD conditional store instruction 426 is used for bounded transactional memory systems, or for other purposes, as described in further detail herein. This instruction tests to see if the property that was set earlier is still present, and if so, stores the value and returns true. Otherwise, a false is returned and the value is not stored. In other words, data is stored at the address only if it has the particular CLMD address that is needed.
  • An example of a conditional store set of hardware instruction definition pseudocode is shown below.
  • whole cache flash CLMD instructions 428 can be provided.
  • a flash clear (AND ALL) instruction 430 can be provided, along with a flash set (OR ALL) instruction 432 .
  • the CLMD_AND_ALL instruction is designed to flash clear all of the CLMDs for each logical processor
  • the CLMD_OR_ALL instruction is designed to flash set all of the CLMDs for each logical processor.
  • FIG. 12 is a diagrammatic view of exemplary context switch save and restore extensions 450 for the central processing unit 200 of FIG. 5 .
  • These are used at context switch time to save and restore the thread context registers (e.g. the architected (programmer visible) thread stage, e.g. various general purpose register files, special registers, etc.).
  • the thread context registers e.g. the architected (programmer visible) thread stage, e.g. various general purpose register files, special registers, etc.
  • both of these instructions take a 512 byte context buffer as an argument.
  • the context switch save and restore instructions can also save the logical processor's overall CMD state structure in some of the context buffer's currently reserved fields.
  • FIGS. 13-16 illustrate some uses of the techniques described herein with a bounded transactional memory application.
  • a bounded transactional memory application can use a programmable subset of the CLMD bits such that if any are ever set on a cache line, indicates that the line has transactionally read (and thus is watched for subsequent writes by other logical processors) or has been speculatively written in a transaction. Being speculatively written means that the line was written without knowledge yet if the values are actually going to be committed permanently or not.
  • the bounded transactional memory application can also include an instruction that atomically tests a subset of the cache metadata CLMD eviction summary bits, and if no such evictions have occurred, atomically clears a subset of the CLMD bit(s) across all cache lines to commit permanently the speculative writes.
  • bounded transactional memory application 470 is one of the application programs that reside on computing device 100 (e.g. is one of hardware assisted software applications 150 ).
  • bounded transactional memory application 470 can alternatively or additionally be embodied as computer-executable instructions on one or more computers and/or in different variations than shown on FIG. 1 .
  • one or more parts of bounded transactional memory application 470 can be part of system memory 104 , on other computers and/or applications 115 , or other such variations as would occur to one in the computer software art.
  • Bounded transactional memory application 470 includes program logic 472 , which is responsible for carrying out some or all of the techniques described herein.
  • Program logic 472 includes logic for accessing the CMD and CLMD metadata in the CPU cache 474 ; logic for setting the CLMD transaction read bit on the address when performing a transactional read 476 ; logic for setting the CLMD transaction write bit on the address and doing a conditional store when performing a transactional write 478 ; logic for testing whether any lines marked as transactional read and write were evicted or invalidated, and if not, flash clearing all the speculative write bits, thereby committing them all atomically 480 ; logic for accessing the metadata to determine whether a transaction is doomed 482 ; and other logic for operating the application 484 .
  • program logic 472 is operable to be called programmatically from another program, such as using a single call to a procedure in program logic 472 .
  • FIG. 14 illustrates one implementation of the stages involved in providing a bounded transactional memory application using cache metadata.
  • the process of FIG. 14 is at least partially implemented in the operating logic of computing device 100 .
  • the process begins at start point 500 with the system providing a bounded transactional memory application with access to CPU cache metadata (stage 502 ).
  • the system sets one of the CLMD bit positions (for example, CLMD[0]), as the CLMD transaction-read bit on the CLMD metadata word for the cache line at that address (stage 504 ).
  • the system For every transactional write, the system sets another of the CLMD bit positions (for example CLMD[1]), as the CLMD transaction-write bit on the CLMD metadata word for the cache line at that address, and does a conditional store (stage 506 ). At commit time, the system tests whether any lines marked as CLMD transaction-read or CLMD transaction-write were evicted or invalidated (stage 508 ). If no evictions/invalidations are found (decision point 510 ), then all the speculative write bits are flash cleared, thereby committing them all (stage 512 ).
  • CLMD[1] the CLMD transaction-write bit on the CLMD metadata word for the cache line at that address
  • This algorithm correctly implements bounded transactional memory for concurrent transactions that that fit in the caches of each logical processor. Since every transactional read of data is explicitly annotated by software with a transaction-read bit on its cache line, and since every speculative transactional write of data is explicitly annotated by software with a transaction-write bit on its cache line, a transaction will only commit if during the time interval of execution, there were no conflicting accesses to data from other logical processors—in particular, if there were no writes (by other logical processors) of data read in the transaction, and if there were no reads (by other logical processors) of data written in the transaction.
  • the behavior of multiprocessor cache coherence is to invalidate this transaction's logical processor's cached copy of the data, an event which due to behavior ( 258 ) is manifest as an eviction/invalidation of non-zero CLMD data in the clmd_evictions register in the CMD ( 190 ). Furthermore, the algorithm correctly watches for any cache line eviction of the entirety of any transactionally accessed data in the time interval from each first reference up to the transaction commit attempt.
  • this event is also manifest via ( 258 ) as an eviction/invalidation in the clmd_evictions register.
  • the combination of software and hardware detects either any transaction conflict or capacity problem and thus guarantees correct transactional memory semantics.
  • FIG. 15 illustrates one implementation of the stages involved in using the CLMD evictions instruction to poll whether a transaction is doomed due to conflicting access or capacity.
  • the process of FIG. 15 is at least partially implemented in the operating logic of computing device 100 .
  • the process begins at start point 530 with repeatedly issuing the GET_CLMD_EVICTIONS instruction to determine whether a transaction is doomed (stage 532 ).
  • An appropriate action is taken to handle the transaction if it is doomed, such as discarding all speculatively written lines and doing a CMD invalidate (stage 534 ).
  • stage 536 ends at end point 536 .
  • FIG. 16 illustrates one implementation of the stages involved in using an addition to the CMD structure to handle doomed transactions in hardware.
  • the process of FIG. 16 is at least partially implemented in the operating logic of computing device 100 .
  • the process begins at start point 550 with initialization of the eviction handler and handler mask (stage 552 ).
  • Some implementations may prevent recursive invocation of the handler by clearing the eviction_handler mask as the handler is first invoked.
  • FIGS. 17-23 a hardware accelerated software transactional memory system that uses cache metadata is described.
  • the cache metadata is used by the software transactional memory system to accelerate some expensive aspects of such a system, such as redundant open_read barrier filtering, redundant write_undo barrier filtering, read log validation, retry operations, and these operations with nested transactions.
  • a TMW describes the transactional synchronization status for associated data that may be accessed in a transaction.
  • the TMW can include a version number, and/or a pointer to a transaction that has the data open for write, and/or a list/count and/or indicator of transactional readers (e.g. pessimistic readers).
  • the list/count and/or indicator of readers can include a count of the number of readers (e.g. pessimistic) accessing the particular value at a given point in time.
  • the list/count and/or indicator of readers can include a list of the particular readers (e.g.
  • the list/count and/or indicator of readers is simply a flag or other indicator to indicate that there are one or more readers (e.g. pessimistic) accessing the particular value at a given point in time.
  • Hardware accelerated software transactional memory application 570 includes program logic 572 , which is responsible for carrying out some or all of the techniques described herein.
  • Program logic 572 includes logic for reserving an opened-for-read bit and a logged-for-undo bit on the VAMD of the CPU cache 574 ; logic for reserving a TMW write-watch bit on the CLMD of the CPU cache 576 ; logic to reset cache metadata between transactions 577 ; logic for providing Open_Read barrier filtering that uses the opened-for-read bit to avoid redundant read logging 578 ; logic for providing Write_Undo barrier filtering that uses the logged-for-undo bit to avoid redundant undo logging 580 ; logic for bypassing read log validation when there have been no read set invalidations (e.g.
  • all compressed metadata state may be quickly reset with a short sequence of instructions such as CLMD_AND_ALL and VAMD_AND_ALL, issued either before the transaction begins or immediately after it ends.
  • FIG. 18 illustrates one implementation of the stages involved in providing Open_Read barrier filtering that uses the opened-for-read bit position on the VAMDs of the cache metadatato efficiently filter out redundant transactional read logging.
  • the process of FIG. 18 is at least partially implemented in the operating logic of computing device 100 .
  • the process begins at start point 600 with the system starting an Open_Read sequence using a hardware accelerated software transactional memory system (stage 602 ). If the VAMD opened-for-read bit is already set for the address of the TMW (decision point 604 ), this indicates this TMW has already been opened for read in this transaction, and software jumps over the read barrier logging sequence (stages 606 , 607 , 608 ).
  • stage 606 sets the opened-for-read bit on the VAMD of the address of the TMW (stage 606 ) and performs the transactional Open_Read logic. In one form, this logs the read access.
  • stage 604 may be implemented with a VAMD_TEST instruction followed by a conditional jump
  • stage 606 may be implemented with a VAMD_SET or VAMD_OR instruction.
  • stage 604 and 606 may together be implemented with a single VAMD_TSET instruction (test then set) followed by a conditional jump.
  • the system also sets a TMW write-watch bit on the CLMD metadata for the TMW's cache line.
  • Stage 607 may be implemented with a CLMD_SET or CLMD_OR instruction. The process ends at end point 612 .
  • FIG. 19 illustrates one implementation of the stages involved in an Open_Write barrier that complements the just described cache metadata based Open_Read filtering.
  • the process of FIG. 19 is at least partially implemented in the operating logic of computing device 100 .
  • the process begins at start point 630 with the system starting an Open_Write sequence using a hardware accelerated software transactional memory system (stage 632 ).
  • opening a TMW for write also grants read access. Therefore the system sets the opened-for-read bit on the VAMD of the TMW using the VAMD_SET instruction (stage 634 ). Then the system performs the transactional Open_Write logic (stage 636 ).
  • the TMW is overwritten with a pointer to this transaction object (indicating it owns the TMW's object(s) data for writing).
  • the TMW is overwritten with a pointer to an entry in a transactional object write log.
  • a bit in the TMW is changed that indicates the TMW is opened for write by a transaction. The process ends at end point 640 .
  • FIG. 20 illustrates one implementation of the stages involved in providing Write_Undo barrier filtering that uses a logged-for-undo bit position in the VAMD cache metadata to efficiently filter out redundant undo logging.
  • the process of FIG. 20 is at least partially implemented in the operating logic of computing device 100 .
  • the process begins at start point 650 with the system starting a write field sequence using a hardware accelerated software transactional memory system (stage 652 ). If the VAMD logged-for-undo bit is already set for the address of the data field to be overwritten (decision point 654 ), then software jumps over the write undo logging sequence (stages 656 , 658 ).
  • stage 654 may be implemented with a VAMD_TEST instruction followed by a conditional jump
  • stage 656 may be implemented with a VAMD_SET or VAMD_OR instruction.
  • stage 654 and stage 656 may together be implemented with a single VAMD_TSET instruction (test then set) followed by a conditional jump. The process ends at end point 660 .
  • FIG. 21 illustrates one implementation of the stages involved in providing read log validation that uses the GET_CLMD_EVICTIONS instruction to bypass read log validation when there have been no read set invalidations (e.g. no conflicting writes from other threads to data this transaction has read).
  • the process of FIG. 21 is at least partially implemented in the operating logic of computing device 100 .
  • the process begins at start point 680 with the system entering commit processing for a transaction using a hardware accelerated software transactional memory system (stage 682 ).
  • stage 682 the system then checks whether any cache lines marked with the CLMD write-watch bit were evicted or invalidated while in this transaction.
  • software issues the GET_CLMD_EVICTIONS instruction to retrieve the evictions summary and tests the write-watch bit; if it is zero, no watched line was evicted or overwritten by another thread. It follows this transaction's read set never suffered a conflicting write access from another thread, and it is safe to skip expensive read log validation. In that case, software conditionally jumps over the read log validation stage (stage 686 ). If the write-watch bit in the evictions summary is set, software performs read log validation (stage 686 ) as usual. In either event, the transactions are committed or otherwise handled as appropriate (stage 692 ). The process ends at end point 692 .
  • FIG. 22 illustrates one implementation of the stages involved in providing retry operations that mark CLMD lines.
  • the process of FIG. 22 is at least partially implemented in the operating logic of computing device 100 .
  • the process begins at start point 700 with determining that a transaction retry operation should be used in a hardware accelerated software transactional memory system (stage 702 ).
  • stage 702 The system rolls back updates to data written in the transaction and releases any data write locks it may hold (stage 704 ).
  • the system then employs the CLMD write-watch bits previously set (stage 607 ) to reduce the overhead of waiting, and improve the latency of resuming, the transaction after another thread updates (writes) the data (stage 706 ).
  • the CLMD cache eviction uses the GET_CLMD_EVICTIONS instruction to interrogate the evictions summary and then test for evictions of CLMDs.
  • software uses the CLMD_SET_EVICTION_HANDLER and CLMD_SET_EVICTION_HANDLER_MASK instructions to initialize the eviction handler to transfer program control to a retry-wakeup handler the moment a cache line with CLMD bit transaction-write is evicted or invalidated (written) by another thread. The thread can then be put to sleep, in one form via a spin-loop or in another by executing a pause instruction to enter a lower power state.
  • stage 708 When any write-watch-metadata annotated line is evicted or invalidated, the system then wakes up the waiting transaction and retries it again (stage 708 ). The process ends at end point 710 .
  • optional stage 705 may also be performed.
  • CLMD_AND_ALL software zeroes the write-watch bit position on all CLMDs in the cache metadata. Software then loops over every read log entry in the read log, re-establishing a write-watch CLMD bit on the cache line of the TMW address found in each read log entry.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Memory System Of A Hierarchy Structure (AREA)
US11/811,370 2005-12-07 2007-06-08 Cache metadata for implementing bounded transactional memory Active 2030-09-14 US8813052B2 (en)

Priority Applications (4)

Application Number Priority Date Filing Date Title
US11/811,370 US8813052B2 (en) 2005-12-07 2007-06-08 Cache metadata for implementing bounded transactional memory
TW097117848A TWI498733B (zh) 2007-06-08 2008-05-15 實作有限交易式記憶體之快取元資料
EP08769911.2A EP2174223B1 (de) 2007-06-08 2008-05-30 Cache-metadaten zum implementieren eines beschränkten transaktionsspeichers
PCT/US2008/065376 WO2008154191A2 (en) 2007-06-08 2008-05-30 Cache metadata for implementing bounded transactional memory

Applications Claiming Priority (3)

Application Number Priority Date Filing Date Title
US74838605P 2005-12-07 2005-12-07
US11/389,451 US8799882B2 (en) 2005-12-07 2006-03-23 Compiler support for optimizing decomposed software transactional memory operations
US11/811,370 US8813052B2 (en) 2005-12-07 2007-06-08 Cache metadata for implementing bounded transactional memory

Related Parent Applications (1)

Application Number Title Priority Date Filing Date
US11/389,451 Continuation-In-Part US8799882B2 (en) 2005-12-07 2006-03-23 Compiler support for optimizing decomposed software transactional memory operations

Publications (2)

Publication Number Publication Date
US20070245099A1 US20070245099A1 (en) 2007-10-18
US8813052B2 true US8813052B2 (en) 2014-08-19

Family

ID=40130616

Family Applications (1)

Application Number Title Priority Date Filing Date
US11/811,370 Active 2030-09-14 US8813052B2 (en) 2005-12-07 2007-06-08 Cache metadata for implementing bounded transactional memory

Country Status (4)

Country Link
US (1) US8813052B2 (de)
EP (1) EP2174223B1 (de)
TW (1) TWI498733B (de)
WO (1) WO2008154191A2 (de)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150040111A1 (en) * 2006-12-28 2015-02-05 Bratin Saha Handling precompiled binaries in a hardware accelerated software transactional memory system
US10061698B2 (en) 2017-01-31 2018-08-28 Qualcomm Incorporated Reducing or avoiding buffering of evicted cache data from an uncompressed cache memory in a compression memory system when stalled write operations occur
US10169181B2 (en) 2016-09-15 2019-01-01 International Business Machines Corporation Efficient validation of transactional memory in a computer processor
US10853247B2 (en) * 2016-03-18 2020-12-01 Industry-Academic Cooperation Foundation, Yonsei University Device for maintaining data consistency between hardware accelerator and host system and method thereof
US20230325191A1 (en) * 2022-04-12 2023-10-12 Microsoft Technology Licensing, Llc Direct swap caching with zero line optimizations

Families Citing this family (96)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8813052B2 (en) * 2005-12-07 2014-08-19 Microsoft Corporation Cache metadata for implementing bounded transactional memory
US20070143755A1 (en) * 2005-12-16 2007-06-21 Intel Corporation Speculative execution past a barrier
US8898652B2 (en) 2006-03-23 2014-11-25 Microsoft Corporation Cache metadata for accelerating software transactional memory
US7978516B2 (en) 2007-12-27 2011-07-12 Pliant Technology, Inc. Flash memory controller having reduced pinout
US8549269B2 (en) * 2009-03-10 2013-10-01 International Business Machines Corporation Method, apparatus or software for processing exceptions produced by an application program
US8195891B2 (en) * 2009-03-30 2012-06-05 Intel Corporation Techniques to perform power fail-safe caching without atomic metadata
US20100332768A1 (en) * 2009-06-26 2010-12-30 Microsoft Corporation Flexible read- and write-monitored and buffered memory blocks
US8370577B2 (en) 2009-06-26 2013-02-05 Microsoft Corporation Metaphysically addressed cache metadata
US8812796B2 (en) * 2009-06-26 2014-08-19 Microsoft Corporation Private memory regions and coherence optimizations
US8161247B2 (en) * 2009-06-26 2012-04-17 Microsoft Corporation Wait loss synchronization
US8356166B2 (en) * 2009-06-26 2013-01-15 Microsoft Corporation Minimizing code duplication in an unbounded transactional memory system by using mode agnostic transactional read and write barriers
US8250331B2 (en) 2009-06-26 2012-08-21 Microsoft Corporation Operating system virtual memory management for hardware transactional memory
US8489864B2 (en) * 2009-06-26 2013-07-16 Microsoft Corporation Performing escape actions in transactions
US8229907B2 (en) * 2009-06-30 2012-07-24 Microsoft Corporation Hardware accelerated transactional memory system with open nested transactions
US8539465B2 (en) 2009-12-15 2013-09-17 Microsoft Corporation Accelerating unbounded memory transactions using nested cache resident transactions
US8533440B2 (en) * 2009-12-15 2013-09-10 Microsoft Corporation Accelerating parallel transactions using cache resident transactions
US8402218B2 (en) * 2009-12-15 2013-03-19 Microsoft Corporation Efficient garbage collection and exception handling in a hardware accelerated transactional memory system
US9092253B2 (en) * 2009-12-15 2015-07-28 Microsoft Technology Licensing, Llc Instrumentation of hardware assisted transactional memory system
US8479053B2 (en) 2010-07-28 2013-07-02 Intel Corporation Processor with last branch record register storing transaction indicator
US8321635B2 (en) * 2010-11-08 2012-11-27 Lsi Corporation Synchronizing commands for preventing data corruption
US9223687B2 (en) * 2012-06-15 2015-12-29 International Business Machines Corporation Determining the logical address of a transaction abort
US9015419B2 (en) 2012-06-15 2015-04-21 International Business Machines Corporation Avoiding aborts due to associativity conflicts in a transactional environment
US9298631B2 (en) 2012-06-15 2016-03-29 International Business Machines Corporation Managing transactional and non-transactional store observability
US9262320B2 (en) * 2012-06-15 2016-02-16 International Business Machines Corporation Tracking transactional execution footprint
US9298469B2 (en) 2012-06-15 2016-03-29 International Business Machines Corporation Management of multiple nested transactions
US9311101B2 (en) 2012-06-15 2016-04-12 International Business Machines Corporation Intra-instructional transaction abort handling
US9298632B2 (en) * 2012-06-28 2016-03-29 Intel Corporation Hybrid cache state and filter tracking of memory operations during a transaction
US9699263B1 (en) 2012-08-17 2017-07-04 Sandisk Technologies Llc. Automatic read and write acceleration of data accessed by virtual machines
US8898430B2 (en) * 2012-12-05 2014-11-25 Arm Limited Fault handling in address translation transactions
US9501398B2 (en) 2012-12-26 2016-11-22 Sandisk Technologies Llc Persistent storage device with NVRAM for staging writes
US9612948B2 (en) 2012-12-27 2017-04-04 Sandisk Technologies Llc Reads and writes between a contiguous data block and noncontiguous sets of logical address blocks in a persistent storage device
US9239751B1 (en) 2012-12-27 2016-01-19 Sandisk Enterprise Ip Llc Compressing data from multiple reads for error control management in memory systems
US9454420B1 (en) 2012-12-31 2016-09-27 Sandisk Technologies Llc Method and system of reading threshold voltage equalization
US9870830B1 (en) 2013-03-14 2018-01-16 Sandisk Technologies Llc Optimal multilevel sensing for reading data from a storage medium
US9367246B2 (en) 2013-03-15 2016-06-14 Sandisk Technologies Inc. Performance optimization of data transfer for soft information generation
US9236886B1 (en) 2013-03-15 2016-01-12 Sandisk Enterprise Ip Llc Universal and reconfigurable QC-LDPC encoder
US9244763B1 (en) 2013-03-15 2016-01-26 Sandisk Enterprise Ip Llc System and method for updating a reading threshold voltage based on symbol transition information
US9159437B2 (en) 2013-06-11 2015-10-13 Sandisk Enterprise IP LLC. Device and method for resolving an LM flag issue
US9384126B1 (en) 2013-07-25 2016-07-05 Sandisk Technologies Inc. Methods and systems to avoid false negative results in bloom filters implemented in non-volatile data storage systems
US9524235B1 (en) 2013-07-25 2016-12-20 Sandisk Technologies Llc Local hash value generation in non-volatile data storage systems
US9361221B1 (en) 2013-08-26 2016-06-07 Sandisk Technologies Inc. Write amplification reduction through reliable writes during garbage collection
US9639463B1 (en) 2013-08-26 2017-05-02 Sandisk Technologies Llc Heuristic aware garbage collection scheme in storage systems
US9298608B2 (en) 2013-10-18 2016-03-29 Sandisk Enterprise Ip Llc Biasing for wear leveling in storage systems
US9442662B2 (en) 2013-10-18 2016-09-13 Sandisk Technologies Llc Device and method for managing die groups
US9436831B2 (en) 2013-10-30 2016-09-06 Sandisk Technologies Llc Secure erase in a memory device
US9263156B2 (en) 2013-11-07 2016-02-16 Sandisk Enterprise Ip Llc System and method for adjusting trip points within a storage device
US9244785B2 (en) 2013-11-13 2016-01-26 Sandisk Enterprise Ip Llc Simulated power failure and data hardening
US9703816B2 (en) 2013-11-19 2017-07-11 Sandisk Technologies Llc Method and system for forward reference logging in a persistent datastore
US9520197B2 (en) 2013-11-22 2016-12-13 Sandisk Technologies Llc Adaptive erase of a storage device
US9520162B2 (en) 2013-11-27 2016-12-13 Sandisk Technologies Llc DIMM device controller supervisor
US9582058B2 (en) 2013-11-29 2017-02-28 Sandisk Technologies Llc Power inrush management of storage devices
US9235245B2 (en) 2013-12-04 2016-01-12 Sandisk Enterprise Ip Llc Startup performance and power isolation
US9703636B2 (en) 2014-03-01 2017-07-11 Sandisk Technologies Llc Firmware reversion trigger and control
US9454448B2 (en) 2014-03-19 2016-09-27 Sandisk Technologies Llc Fault testing in storage devices
US9390814B2 (en) 2014-03-19 2016-07-12 Sandisk Technologies Llc Fault detection and prediction for data storage elements
US9448876B2 (en) 2014-03-19 2016-09-20 Sandisk Technologies Llc Fault detection and prediction in storage devices
US9971627B2 (en) 2014-03-26 2018-05-15 Intel Corporation Enabling maximum concurrency in a hybrid transactional memory system
US9626399B2 (en) 2014-03-31 2017-04-18 Sandisk Technologies Llc Conditional updates for reducing frequency of data modification operations
US9390021B2 (en) 2014-03-31 2016-07-12 Sandisk Technologies Llc Efficient cache utilization in a tiered data structure
US9626400B2 (en) 2014-03-31 2017-04-18 Sandisk Technologies Llc Compaction of information in tiered data structure
US9697267B2 (en) 2014-04-03 2017-07-04 Sandisk Technologies Llc Methods and systems for performing efficient snapshots in tiered data structures
US10162748B2 (en) 2014-05-30 2018-12-25 Sandisk Technologies Llc Prioritizing garbage collection and block allocation based on I/O history for logical address regions
US10114557B2 (en) 2014-05-30 2018-10-30 Sandisk Technologies Llc Identification of hot regions to enhance performance and endurance of a non-volatile storage device
US10146448B2 (en) 2014-05-30 2018-12-04 Sandisk Technologies Llc Using history of I/O sequences to trigger cached read ahead in a non-volatile storage device
US9703491B2 (en) 2014-05-30 2017-07-11 Sandisk Technologies Llc Using history of unaligned writes to cache data and avoid read-modify-writes in a non-volatile storage device
US10656842B2 (en) 2014-05-30 2020-05-19 Sandisk Technologies Llc Using history of I/O sizes and I/O sequences to trigger coalesced writes in a non-volatile storage device
US10372613B2 (en) 2014-05-30 2019-08-06 Sandisk Technologies Llc Using sub-region I/O history to cache repeatedly accessed sub-regions in a non-volatile storage device
US10656840B2 (en) 2014-05-30 2020-05-19 Sandisk Technologies Llc Real-time I/O pattern recognition to enhance performance and endurance of a storage device
US9652381B2 (en) 2014-06-19 2017-05-16 Sandisk Technologies Llc Sub-block garbage collection
US9720837B2 (en) 2014-06-27 2017-08-01 International Business Machines Corporation Allowing non-cacheable loads within a transaction
US9658961B2 (en) 2014-06-27 2017-05-23 International Business Machines Corporation Speculation control for improving transaction success rate, and instruction therefor
US9740614B2 (en) 2014-06-27 2017-08-22 International Business Machines Corporation Processor directly storing address range of co-processor memory accesses in a transactional memory where co-processor supplements functions of the processor
US9703718B2 (en) 2014-06-27 2017-07-11 International Business Machines Corporation Managing read tags in a transactional memory
US10025715B2 (en) 2014-06-27 2018-07-17 International Business Machines Corporation Conditional inclusion of data in a transactional memory read set
US9772944B2 (en) 2014-06-27 2017-09-26 International Business Machines Corporation Transactional execution in a multi-processor environment that monitors memory conflicts in a shared cache
US10073784B2 (en) 2014-06-27 2018-09-11 International Business Machines Corporation Memory performance when speculation control is enabled, and instruction therefor
US9477481B2 (en) 2014-06-27 2016-10-25 International Business Machines Corporation Accurate tracking of transactional read and write sets with speculation
US10114752B2 (en) 2014-06-27 2018-10-30 International Business Machines Corporation Detecting cache conflicts by utilizing logical address comparisons in a transactional memory
US10013351B2 (en) 2014-06-27 2018-07-03 International Business Machines Corporation Transactional execution processor having a co-processor accelerator, both sharing a higher level cache
GB2528115B (en) * 2014-07-11 2021-05-19 Advanced Risc Mach Ltd Dynamic saving of registers in transactions
US9443601B2 (en) 2014-09-08 2016-09-13 Sandisk Technologies Llc Holdup capacitor energy harvesting
US10740108B2 (en) 2017-04-18 2020-08-11 International Business Machines Corporation Management of store queue based on restoration operation
US10552164B2 (en) 2017-04-18 2020-02-04 International Business Machines Corporation Sharing snapshots between restoration and recovery
US10572265B2 (en) 2017-04-18 2020-02-25 International Business Machines Corporation Selecting register restoration or register reloading
US10649785B2 (en) 2017-04-18 2020-05-12 International Business Machines Corporation Tracking changes to memory via check and recovery
US10545766B2 (en) 2017-04-18 2020-01-28 International Business Machines Corporation Register restoration using transactional memory register snapshots
US10963261B2 (en) 2017-04-18 2021-03-30 International Business Machines Corporation Sharing snapshots across save requests
US10838733B2 (en) 2017-04-18 2020-11-17 International Business Machines Corporation Register context restoration based on rename register recovery
US10782979B2 (en) 2017-04-18 2020-09-22 International Business Machines Corporation Restoring saved architected registers and suppressing verification of registers to be restored
US11010192B2 (en) 2017-04-18 2021-05-18 International Business Machines Corporation Register restoration using recovery buffers
US12135655B2 (en) 2017-07-27 2024-11-05 International Business Machines Corporation Saving track metadata format information for tracks demoted from cache for use when the demoted track is later staged into cache
US10691566B2 (en) 2017-07-27 2020-06-23 International Business Machines Corporation Using a track format code in a cache control block for a track in a cache to process read and write requests to the track in the cache
US11036641B2 (en) 2017-08-09 2021-06-15 International Business Machines Corporation Invalidating track format information for tracks demoted from cache
US10579532B2 (en) 2017-08-09 2020-03-03 International Business Machines Corporation Invalidating track format information for tracks in cache
US10565109B2 (en) * 2017-09-05 2020-02-18 International Business Machines Corporation Asynchronous update of metadata tracks in response to a cache hit generated via an I/O operation over a bus interface
US10942677B2 (en) * 2019-01-24 2021-03-09 Silicon Motion, Inc. Method for performing access management of memory device, associated memory device and controller thereof, associated host device and associated electronic device

Citations (46)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5428761A (en) * 1992-03-12 1995-06-27 Digital Equipment Corporation System for achieving atomic non-sequential multi-word operations in shared memory
US5864849A (en) 1996-12-16 1999-01-26 Lucent Technologies Inc. System and method for restoring a multiple checkpointed database in view of loss of volatile memory
US5940838A (en) 1997-07-11 1999-08-17 International Business Machines Corporation Parallel file system and method anticipating cache usage patterns
US6081883A (en) 1997-12-05 2000-06-27 Auspex Systems, Incorporated Processing system with dynamically allocatable buffer memory
US6230311B1 (en) * 1998-06-12 2001-05-08 International Business Machines Corporation Apparatus and method for disabling methods called on an object
US6266742B1 (en) 1997-10-27 2001-07-24 International Business Machines Corporation Algorithm for cache replacement
US6327175B1 (en) 1999-09-13 2001-12-04 Cypress Semiconductor Corporation Method and apparatus for controlling a memory array with a programmable register
US20020083353A1 (en) 2000-12-07 2002-06-27 Doron Orenstein Low-power processor hint, such as from a PAUSE instruction
US20020087614A1 (en) 2000-08-31 2002-07-04 Andrej Kocev Programmable tuning for flow control and support for CPU hot plug
US20020091718A1 (en) 1998-09-04 2002-07-11 Philip L. Bohannon Method and apparatus for detecting and recovering from data corruption of database via read logging
US6571259B1 (en) 2000-09-26 2003-05-27 Emc Corporation Preallocation of file system cache blocks in a data storage system
US6658539B2 (en) * 2001-10-16 2003-12-02 International Business Machines Corporation Super-coherent data mechanisms for shared caches in a multiprocessing system
US20040015642A1 (en) 2002-07-16 2004-01-22 Sun Microsystems, Inc. Software transactional memory for dynamically sizable shared data structures
WO2004042583A2 (en) 2002-11-01 2004-05-21 Bluearc Uk Limited Apparatus and method for hardware-based file system
US6754773B2 (en) 2001-01-29 2004-06-22 Snap Appliance, Inc. Data engine with metadata processor
US20040133769A1 (en) * 2002-12-24 2004-07-08 Shailender Chaudhry Generating prefetches by speculatively executing code through hardware scout threading
US6775727B2 (en) 2001-06-23 2004-08-10 Freescale Semiconductor, Inc. System and method for controlling bus arbitration during cache memory burst cycles
US20040162968A1 (en) 2003-02-13 2004-08-19 Marc Tremblay Fail instruction to support transactional program execution
US20040163082A1 (en) * 2003-02-13 2004-08-19 Marc Tremblay Commit instruction to support transactional program execution
US20040187116A1 (en) * 2003-02-13 2004-09-23 Marc Tremblay Selectively monitoring loads to support transactional program execution
US20040187123A1 (en) * 2003-02-13 2004-09-23 Marc Tremblay Selectively unmarking load-marked cache lines during transactional program execution
US20050060383A1 (en) 2003-09-12 2005-03-17 Lovett Thomas D. Temporary storage of memory line while waiting for cache eviction
US6880045B2 (en) 1999-02-26 2005-04-12 Hewlett-Packard Development Company, L.P. Multi-processor computer system with transactional memory
US20050086446A1 (en) 2003-10-04 2005-04-21 Mckenney Paul E. Utilizing software locking approach to execute code upon failure of hardware transactional approach
US20050120179A1 (en) * 2003-12-02 2005-06-02 Intel Corporation (A Delaware Corporation) Single-version data cache with multiple checkpoint support
US20050210198A1 (en) 2004-03-22 2005-09-22 International Business Machines Corporation Method and apparatus for prefetching data from a data structure
US20060010227A1 (en) * 2004-06-01 2006-01-12 Rajeev Atluri Methods and apparatus for accessing data from a primary data storage system for secondary storage
US20060085591A1 (en) 2004-09-30 2006-04-20 Sanjeev Kumar Hybrid hardware and software implementation of transactional memory access
US20060161740A1 (en) 2004-12-29 2006-07-20 Sailesh Kottapalli Transaction based shared data operations in a multiprocessor environment
US7114036B2 (en) 2004-01-14 2006-09-26 International Business Machines Corporation Method and apparatus for autonomically moving cache entries to dedicated storage when false cache line sharing is detected
US7181608B2 (en) 2000-02-03 2007-02-20 Realtime Data Llc Systems and methods for accelerated loading of operating systems and application programs
US7188216B1 (en) 2002-12-13 2007-03-06 Vignette Corporation Method and system for an extensible caching framework
US20070073528A1 (en) 2005-09-28 2007-03-29 William Watt Hardware acceleration system for logic simulation using shift register as local cache
US7206903B1 (en) * 2004-07-20 2007-04-17 Sun Microsystems, Inc. Method and apparatus for releasing memory locations during transactional execution
US20070143287A1 (en) 2005-12-15 2007-06-21 Ali-Reza Adl-Tabatabai Coordinating access to memory locations for hardware transactional memory transactions and software transactional memory transactions
US20070186056A1 (en) 2006-02-07 2007-08-09 Bratin Saha Hardware acceleration for a software transactional memory system
US20070239942A1 (en) * 2006-03-30 2007-10-11 Ravi Rajwar Transactional memory virtualization
US20070245099A1 (en) * 2005-12-07 2007-10-18 Microsoft Corporation Cache metadata for implementing bounded transactional memory
US20070245128A1 (en) * 2006-03-23 2007-10-18 Microsoft Corporation Cache metadata for accelerating software transactional memory
US20070245309A1 (en) * 2005-12-07 2007-10-18 Microsoft Corporation Software accessible cache metadata
US20070260942A1 (en) * 2006-03-30 2007-11-08 Ravi Rajwar Transactional memory in out-of-order processors
US20080010417A1 (en) * 2006-04-28 2008-01-10 Zeffer Hakan E Read/Write Permission Bit Support for Efficient Hardware to Software Handover
US20080040551A1 (en) * 2005-12-07 2008-02-14 Microsoft Corporation Cache metadata identifiers for isolation and sharing
US20080162990A1 (en) * 2006-12-28 2008-07-03 Cheng Wang Compiler technique for efficient register checkpointing to support transaction roll-back
US20080163220A1 (en) * 2006-12-28 2008-07-03 Cheng Wang Efficient and consistent software transactional memory
US7822727B1 (en) 2004-07-02 2010-10-26 Borland Software Corporation System and methodology for performing read-only transactions in a shared cache

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWM309149U (en) * 2006-06-23 2007-04-01 Genesys Logic Inc Data cache device of flash memory

Patent Citations (51)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5428761A (en) * 1992-03-12 1995-06-27 Digital Equipment Corporation System for achieving atomic non-sequential multi-word operations in shared memory
US5864849A (en) 1996-12-16 1999-01-26 Lucent Technologies Inc. System and method for restoring a multiple checkpointed database in view of loss of volatile memory
US5940838A (en) 1997-07-11 1999-08-17 International Business Machines Corporation Parallel file system and method anticipating cache usage patterns
US6266742B1 (en) 1997-10-27 2001-07-24 International Business Machines Corporation Algorithm for cache replacement
US6081883A (en) 1997-12-05 2000-06-27 Auspex Systems, Incorporated Processing system with dynamically allocatable buffer memory
US6230311B1 (en) * 1998-06-12 2001-05-08 International Business Machines Corporation Apparatus and method for disabling methods called on an object
US20020091718A1 (en) 1998-09-04 2002-07-11 Philip L. Bohannon Method and apparatus for detecting and recovering from data corruption of database via read logging
US6449623B1 (en) 1998-09-04 2002-09-10 Lucent Technologies Inc, Method and apparatus for detecting and recovering from data corruption of a database via read logging
US6880045B2 (en) 1999-02-26 2005-04-12 Hewlett-Packard Development Company, L.P. Multi-processor computer system with transactional memory
US6327175B1 (en) 1999-09-13 2001-12-04 Cypress Semiconductor Corporation Method and apparatus for controlling a memory array with a programmable register
US7181608B2 (en) 2000-02-03 2007-02-20 Realtime Data Llc Systems and methods for accelerated loading of operating systems and application programs
US20020087614A1 (en) 2000-08-31 2002-07-04 Andrej Kocev Programmable tuning for flow control and support for CPU hot plug
US6571259B1 (en) 2000-09-26 2003-05-27 Emc Corporation Preallocation of file system cache blocks in a data storage system
US20020083353A1 (en) 2000-12-07 2002-06-27 Doron Orenstein Low-power processor hint, such as from a PAUSE instruction
US6754773B2 (en) 2001-01-29 2004-06-22 Snap Appliance, Inc. Data engine with metadata processor
US6775727B2 (en) 2001-06-23 2004-08-10 Freescale Semiconductor, Inc. System and method for controlling bus arbitration during cache memory burst cycles
US6658539B2 (en) * 2001-10-16 2003-12-02 International Business Machines Corporation Super-coherent data mechanisms for shared caches in a multiprocessing system
US20040015642A1 (en) 2002-07-16 2004-01-22 Sun Microsystems, Inc. Software transactional memory for dynamically sizable shared data structures
WO2004042583A2 (en) 2002-11-01 2004-05-21 Bluearc Uk Limited Apparatus and method for hardware-based file system
US7188216B1 (en) 2002-12-13 2007-03-06 Vignette Corporation Method and system for an extensible caching framework
US20040133769A1 (en) * 2002-12-24 2004-07-08 Shailender Chaudhry Generating prefetches by speculatively executing code through hardware scout threading
US20040162968A1 (en) 2003-02-13 2004-08-19 Marc Tremblay Fail instruction to support transactional program execution
US20040163082A1 (en) * 2003-02-13 2004-08-19 Marc Tremblay Commit instruction to support transactional program execution
US20040187116A1 (en) * 2003-02-13 2004-09-23 Marc Tremblay Selectively monitoring loads to support transactional program execution
US20040187123A1 (en) * 2003-02-13 2004-09-23 Marc Tremblay Selectively unmarking load-marked cache lines during transactional program execution
US20050060383A1 (en) 2003-09-12 2005-03-17 Lovett Thomas D. Temporary storage of memory line while waiting for cache eviction
US20050086446A1 (en) 2003-10-04 2005-04-21 Mckenney Paul E. Utilizing software locking approach to execute code upon failure of hardware transactional approach
US20050120179A1 (en) * 2003-12-02 2005-06-02 Intel Corporation (A Delaware Corporation) Single-version data cache with multiple checkpoint support
US7114036B2 (en) 2004-01-14 2006-09-26 International Business Machines Corporation Method and apparatus for autonomically moving cache entries to dedicated storage when false cache line sharing is detected
US20050210198A1 (en) 2004-03-22 2005-09-22 International Business Machines Corporation Method and apparatus for prefetching data from a data structure
US20060010227A1 (en) * 2004-06-01 2006-01-12 Rajeev Atluri Methods and apparatus for accessing data from a primary data storage system for secondary storage
US7822727B1 (en) 2004-07-02 2010-10-26 Borland Software Corporation System and methodology for performing read-only transactions in a shared cache
US7206903B1 (en) * 2004-07-20 2007-04-17 Sun Microsystems, Inc. Method and apparatus for releasing memory locations during transactional execution
US20060085591A1 (en) 2004-09-30 2006-04-20 Sanjeev Kumar Hybrid hardware and software implementation of transactional memory access
US20060161740A1 (en) 2004-12-29 2006-07-20 Sailesh Kottapalli Transaction based shared data operations in a multiprocessor environment
US20070073528A1 (en) 2005-09-28 2007-03-29 William Watt Hardware acceleration system for logic simulation using shift register as local cache
US20070245099A1 (en) * 2005-12-07 2007-10-18 Microsoft Corporation Cache metadata for implementing bounded transactional memory
US20070245309A1 (en) * 2005-12-07 2007-10-18 Microsoft Corporation Software accessible cache metadata
US20080040551A1 (en) * 2005-12-07 2008-02-14 Microsoft Corporation Cache metadata identifiers for isolation and sharing
US8225297B2 (en) 2005-12-07 2012-07-17 Microsoft Corporation Cache metadata identifiers for isolation and sharing
US8001538B2 (en) 2005-12-07 2011-08-16 Microsoft Corporation Software accessible cache metadata
US20070143287A1 (en) 2005-12-15 2007-06-21 Ali-Reza Adl-Tabatabai Coordinating access to memory locations for hardware transactional memory transactions and software transactional memory transactions
US20070186056A1 (en) 2006-02-07 2007-08-09 Bratin Saha Hardware acceleration for a software transactional memory system
WO2007092422A2 (en) 2006-02-07 2007-08-16 Intel Corporation Hardware acceleration for a software transactional memory system
US20080046661A1 (en) 2006-02-07 2008-02-21 Bratin Saha Hardware acceleration for a software transactional memory system
US20070245128A1 (en) * 2006-03-23 2007-10-18 Microsoft Corporation Cache metadata for accelerating software transactional memory
US20070239942A1 (en) * 2006-03-30 2007-10-11 Ravi Rajwar Transactional memory virtualization
US20070260942A1 (en) * 2006-03-30 2007-11-08 Ravi Rajwar Transactional memory in out-of-order processors
US20080010417A1 (en) * 2006-04-28 2008-01-10 Zeffer Hakan E Read/Write Permission Bit Support for Efficient Hardware to Software Handover
US20080163220A1 (en) * 2006-12-28 2008-07-03 Cheng Wang Efficient and consistent software transactional memory
US20080162990A1 (en) * 2006-12-28 2008-07-03 Cheng Wang Compiler technique for efficient register checkpointing to support transaction roll-back

Non-Patent Citations (14)

* Cited by examiner, † Cited by third party
Title
Adl-Tabatabai, Ali-Reza, et al. "Unlocking Concurrency." ACM Queue.10 pages. Dec./Jan. 2006-2007.
Chuang et al., "Unbounded Page-Based Transactional Memory," ASPLOS'06, Oct. 21-25, 2006, San Jose, CA, copyright 2006 ACM 12 pages.
European Search Report dated Sep. 4, 2012, European Patent Application No. 08 769 911.2, 14 pages.
Harris, Tim, et al. "Optimizing Memory Transactions." 12 pages. 2006.
International Search Report and Written Opinion for International Application No. PCT/US2008/065381, Jan. 16, 2009, 11 pages.
International Search Report and Written Opinion Received for PCT Application No. PCT/US2008/065370, mailed on Oct. 31, 2008, 6 pages.
International Search Report and Written Opinion Received for PCT Application No. PCT/US2008/065376, mailed on Dec. 31, 2008, 10 pages.
Marathe, Virendra J., et al. "Lowering the Overhead of Nonblocking Sortware Transactional Memory." 11 pages. May 17, 2006.
Moore et al., "LogTM: Log-based Transactional Memory," The 12th International Symposium on High-Performance Computer Architecture, Jan. 1, 2006, pp. 258-269.
Moore, Kevin, et al. "LogTM: Log-based Transactional Memory." 12th Annual International Symposium on High Performance Computer Architecture (HPCA-12). 12 pages. Feb. 2006.
Rajwar, et al., "Virtualizing Transactional Memory," Proceedings of the 32nd International Symposium on Computer Architecture (ISCA '05), copyright 2005 IEEE, 12 pages (Jun. 2005).
Saha, Bratin, et al. "Architectural Support for Software Transactional Memory." The 39th Annual IEEE/ACM International Symposium on Microarchitecture (MICRO'06). 12 pages. 2006.
Search Report for Taiwan issued in Patent Application No. 097117848, dated Oct. 15, 2013, 1 page.
Shriraman, et al., "Hardware Acceleration of Software Transactional Memory," The First ACM SIGPLAN 1 Workshop on Languages, Compilers, and Hardware Support for Transactional Computing, held in conjunction with PLDI, Ottawa, Ontario, Canada, Jun. 2006, 10 pages.

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150040111A1 (en) * 2006-12-28 2015-02-05 Bratin Saha Handling precompiled binaries in a hardware accelerated software transactional memory system
US9304769B2 (en) * 2006-12-28 2016-04-05 Intel Corporation Handling precompiled binaries in a hardware accelerated software transactional memory system
US10853247B2 (en) * 2016-03-18 2020-12-01 Industry-Academic Cooperation Foundation, Yonsei University Device for maintaining data consistency between hardware accelerator and host system and method thereof
US10169181B2 (en) 2016-09-15 2019-01-01 International Business Machines Corporation Efficient validation of transactional memory in a computer processor
US10061698B2 (en) 2017-01-31 2018-08-28 Qualcomm Incorporated Reducing or avoiding buffering of evicted cache data from an uncompressed cache memory in a compression memory system when stalled write operations occur
US20230325191A1 (en) * 2022-04-12 2023-10-12 Microsoft Technology Licensing, Llc Direct swap caching with zero line optimizations
US11847459B2 (en) * 2022-04-12 2023-12-19 Microsoft Technology Licensing, Llc Direct swap caching with zero line optimizations
US12204909B2 (en) * 2022-04-12 2025-01-21 Microsoft Technology Licensing, Llc Direct swap caching with zero line optimizations

Also Published As

Publication number Publication date
EP2174223B1 (de) 2019-05-15
WO2008154191A2 (en) 2008-12-18
TW200905474A (en) 2009-02-01
TWI498733B (zh) 2015-09-01
US20070245099A1 (en) 2007-10-18
WO2008154191A3 (en) 2009-02-19
EP2174223A2 (de) 2010-04-14
EP2174223A4 (de) 2012-10-03

Similar Documents

Publication Publication Date Title
US8813052B2 (en) Cache metadata for implementing bounded transactional memory
US8898652B2 (en) Cache metadata for accelerating software transactional memory
US8001538B2 (en) Software accessible cache metadata
US8225297B2 (en) Cache metadata identifiers for isolation and sharing
US8229907B2 (en) Hardware accelerated transactional memory system with open nested transactions
US8356166B2 (en) Minimizing code duplication in an unbounded transactional memory system by using mode agnostic transactional read and write barriers
US9329890B2 (en) Managing high-coherence-miss cache lines in multi-processor computing environments
US10235201B2 (en) Dynamic releasing of cache lines
US20110208921A1 (en) Inverted default semantics for in-speculative-region memory accesses
US9952976B2 (en) Allowing non-cacheable loads within a transaction
US10884931B2 (en) Interprocessor memory status communication
US20100332768A1 (en) Flexible read- and write-monitored and buffered memory blocks
US9921872B2 (en) Interprocessor memory status communication
US10261828B2 (en) Interprocessor memory status communication
US9916180B2 (en) Interprocessor memory status communication

Legal Events

Date Code Title Description
AS Assignment

Owner name: MICROSOFT CORPORATION, WASHINGTON

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:GRAY, JAN;HARRIS, TIMOTHY L.;LARUS, JAMES;AND OTHERS;REEL/FRAME:020265/0747

Effective date: 20070608

FEPP Fee payment procedure

Free format text: PAYOR NUMBER ASSIGNED (ORIGINAL EVENT CODE: ASPN); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

STCF Information on status: patent grant

Free format text: PATENTED CASE

AS Assignment

Owner name: MICROSOFT TECHNOLOGY LICENSING, LLC, WASHINGTON

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:MICROSOFT CORPORATION;REEL/FRAME:034542/0001

Effective date: 20141014

MAFP Maintenance fee payment

Free format text: PAYMENT OF MAINTENANCE FEE, 4TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1551)

Year of fee payment: 4

MAFP Maintenance fee payment

Free format text: PAYMENT OF MAINTENANCE FEE, 8TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1552); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

Year of fee payment: 8

MAFP Maintenance fee payment

Free format text: PAYMENT OF MAINTENANCE FEE, 12TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1553); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

Year of fee payment: 12